Skip to content

Instantly share code, notes, and snippets.

@mattkretz
Created October 1, 2018 12:41
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save mattkretz/af3bf180606379155f175a8ed8b187de to your computer and use it in GitHub Desktop.
Save mattkretz/af3bf180606379155f175a8ed8b187de to your computer and use it in GitHub Desktop.
Changelog Vc 1.4.0
  • 45a75e0a Adjust to new version.h path and fix commit - Matthias Kretz
  • 6c6b16c2 Document make_interleave_wrapper - Matthias Kretz
  • 98cc3c90 Use correct accessor for const memory - Matthias Kretz
  • af621f1f Rewrite operator return type deduction - Matthias Kretz
  • e519eae8 Test operator failures via sfinae_is_callable - Matthias Kretz
  • c9bba807 MSVC: work around C++ frontend bug - Matthias Kretz
  • 41c21fbb ICC uses movups, which doesn't break the ABI - Matthias Kretz
  • 67dc8e3b Fix workaround & disambiguate fixed_size simd_cast - Matthias Kretz
  • 9d9ad90d Move enable_if to TPL to unbreak clang 3.8 - Matthias Kretz
  • 60d2637b Remove many static_assert sanity-checks - Matthias Kretz
  • 11a93f02 * Use fixed_size_simd<int, Size> as IndexType for all ABIs * Define integral AVX compares only when AVX2 is available * Refactor binary operators to handle (mixed) fixed_size_simd(_mask) and Simd(Mask)Array better (or at all) * Modify Simd(Mask)Array operations to return fixed_size_simd(_mask) * Refactor some enable_ifs to move from default arg to template parameter list. This is especially important to work around constructor inheritance bugs in GCC. * Make sure fixed_size unconditionally passes via the stack (to solve potential ABI breakage on linking TUs compiled with different -march flags). * Disallow implicit narrowing conversions on binary operators (e.g. simd + long) - Matthias Kretz
  • 9589fce4 Simplify Traits::vector_size - Matthias Kretz
  • 7b4327aa Enforce SimdArray template arguments as intended - Matthias Kretz
  • b45acb66 Dox: fix search and hide Vc_VDECL - Matthias Kretz
  • d4faa212 AppVeyor: use MSVC2017 and init git submodules - Matthias Kretz
  • dcef14be GCC: Optimize constant folding of compares - Matthias Kretz
  • dc689f5a Auto-publish release documentation - Matthias Kretz
  • ccd40973 Mention git submodules, link 1.4 docs, fix linebreak - Matthias Kretz
  • 138afe90 MSVC: work around overload resolution confusion - Matthias Kretz
  • b120d86c MSVC: fix isnan(fixed_size_simd) somehow returning bool - Matthias Kretz
  • 510396e4 MSVC: drop previous workaround after MSVC broke compat - Matthias Kretz
  • 71a43323 Win10/MSVC doesn't fail on std::fma anymore - Matthias Kretz
  • bb2fb1ef Make operators less ambiguous (fixed_size_simd vs. SimdArray) - Matthias Kretz
  • ada89514 Bump virtest for MSVC workarounds - Matthias Kretz
  • b8835299 Fix enable_if overload disambiguation without packs - Matthias Kretz
  • 9ea6a79b Bump double fuzzyness for sincos - Matthias Kretz
  • 2197011c Travis: Bump MacOS GCC versions and allow them to fail - Matthias Kretz
  • 774317ca Bump virtest for the fixes in it - Matthias Kretz
  • 358e906b Add missing include for std::size_t - Matthias Kretz
  • 6c108c66 Remove more MIC mentions - Matthias Kretz
  • c0f2b288 CTest: test "make install/fast" - Matthias Kretz
  • 37c277b8 CTest: Quick exit for ICC<18 & less useless error codes - Matthias Kretz
  • d2bff6fc Fix test object construction - Matthias Kretz
  • 42c75d8b Partial revert of the MemoryVector change - Matthias Kretz
  • 5293707f Fix includes after director reorganization - Matthias Kretz
  • f23184be Add another Compiler Explorer example - Matthias Kretz
  • 8876cadc Adapt to changed directory (include/Vc -> Vc) - Dennis Klein
  • d5e6094a Fix copy&paste bug - Dennis Klein
  • c04b3373 Add and test generator ctor for simdize - Matthias Kretz
  • 4accb4cc Update godbolt header to latest src - Matthias Kretz
  • dec1cfc6 Replace reinterpret_cast<MayAlias*> with aliasing_cast - Matthias Kretz
  • 43a65148 * Optimize 3-arg (de)interleave from successive indexes with AVX * New load_interleaved and store_interleaved functions for simdized objects. This uses existing interleave code, when the sizeof all members is equal (homogeneous_sizeof). Otherwise it'll use loads from scalars and depend on the compiler to do something smarter. * Modify simd_for_each(_n) to use simdize<value_type> and thus vectorize everything that's thrown at it. This uses the new load_interleaved and store_interleaved functions to improve the chunked AoS <-> AoVS conversions. - Matthias Kretz
  • 520d8b54 Add link to Compiler Explorer example - Matthias Kretz
  • 067196f2 Drop another MIC remnant - Matthias Kretz
  • bc6e9333 Make simdize<T, N> smarter, preferring native ABI - Matthias Kretz
  • ef2625ae Optimize several AVX (de)interleaves - Matthias Kretz
  • c78174fb Fix relative include to use quotes - Matthias Kretz
  • c10c210b Doxygen cleanup - Matthias Kretz
  • a97edffb Clean up include and declaration order - Matthias Kretz
  • fa6a2cd1 Modernize CMake package and add exported targets - Dennis Klein
  • 4b785cde Update single header after include fixes - Matthias Kretz
  • 1efc1877 CMake: Fix installation - Matthias Kretz
  • 942c7875 Remove deprecated headers - Matthias Kretz
  • 5519ed06 Reorganize directories to match installation - Matthias Kretz
  • e1c4bd6e Single header for godbolt.org - Matthias Kretz
  • da155c3c Drop Open64 remnant - Matthias Kretz
  • bd30b199 Drop all MIC code - Matthias Kretz
  • 2ab99c8b Add simple fpclassify, mostly for virtest's ulp code - Matthias Kretz
  • 6b77615d Port to latest virtest; add some std::simd compat - Matthias Kretz
  • 18948403 Bump version to 1.3.80-dev - Matthias Kretz
  • 9aacbc0c Port remaining test to new test interface - Matthias Kretz
  • 246ca17c Improve documentation on how to work with containers - Matthias Kretz
  • fa9cf783 Avoid aliasing issue; fixes GCC8 test failure - Matthias Kretz
  • b064211e Fix math reference compilation for MIC - Matthias Kretz
  • 78e329f3 ICC also doesn't like the constexpr in span - Matthias Kretz
  • f30a25a1 Fix MIC gather to work with Segments - Matthias Kretz
  • 20bef274 Optimize & avoid miscompilation simple loop gather on GCC - Matthias Kretz
  • ccbb0501 Add fallthrough comment for GCC 7+ - Matthias Kretz
  • 8bdc4e53 Add 'default: unreachable' to switch - Matthias Kretz
  • da82e215 Rename reference to x to avoid shadowing - Matthias Kretz
  • bb718cd6 Standalone toplevel Makefile - Matthias Kretz
  • e801c348 Add a cmake option to use libmvec for sin & cos - Matthias Kretz
  • a44409e4 Use a sane stack-protector for libVc - Matthias Kretz
  • 5d32f356 Use testdata from vc-testdata git submodule - Matthias Kretz
  • 368725ed Refactor & fix sin/cos for more precision - Matthias Kretz
  • 46073e4a Octave, bc, and gnuplot scripts supporting math dev - Matthias Kretz
  • e2f53519 Document containers (esp. Vc::span) - Matthias Kretz
  • 7182f0d7 Add Vc::span, supporting gather & scatter - Matthias Kretz
  • f6b63aa9 Use T/F instead of True/False in conditional - Arif Ahmed
  • d8ba732d Enable better optimization with GCC - Matthias Kretz
  • 288d0d92 Fix non-SSE4 trunc, floor, and ceil - Matthias Kretz
  • de755804 Remove unused FindSSE.cmake file - Matthias Kretz
  • 54a697e4 CMake: Add convenience toplevel Makefile - Matthias Kretz
  • 5d5cb248 CMake: Default USE_LIBC++ to OFF - Matthias Kretz
  • c7143d92 CTest: Make sure $CXX invocation doesn't quote spaces - Matthias Kretz
  • ac2b9501 CTest: drop old warning suppression - Matthias Kretz
  • a59f5cc7 On OSX the name of the global var is different - Matthias Kretz
  • 2f3cc158 Travis: No default GCC on OSX anymore - Matthias Kretz
  • 25505f40 Support new ABI check with ICC/MIC - Matthias Kretz
  • 71a81712 Clang: make the ABI test work - Matthias Kretz
  • db74285b Travis: work around unfriendly defaults on MacOS - Matthias Kretz
  • 05aaee53 Improve codegen on argument passing - Matthias Kretz
  • 721f53f5 Rewrite ABI test logic Refs: gh-191 - Matthias Kretz
  • 53e12bd0 For architectures without optimized implementation use scalar one - StefanBruens
  • b8835e1e Implemented #182, i.e. GCC specific workaround for std::for_each_n. - Tom Westerhout
  • 13ab654f Speedup CPU detection on mac - Giulio Eulisse
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment